Large-scale proteomics turnaround with Artificial Intelligence
Analyzing large-scale proteomics datasets helps uncover biomarkers for neurodegenerative diseases. AI and multi-task learning improve predictive insights.
Challenge
A biotech company needed to infer biomarkers indicative of neurodegenerative diseases using Olink proteomics assays on a large-scale dataset.
Approach
- Designed a machine learning model capable of multi-task predictions across 400+ indications.
- Trained the model on 50,000+ patients and 150 million data points.
- Focused on identifying robust protein signatures and patient stratification.
Results
- Identified biomarkers validated by orthogonal studies.
- Improved baseline benchmarks by 16%, with success in 91% of tasks.
This AI-led approach unlocks critical biological insights, supporting the discovery of biomarkers for neurodegenerative diseases.
Expert Contribution
Reviewed by: Dr. Krzysztof Kolmus, PhD
Role: Lead Bioinformatician, AI‑Driven Drug Discovery
Expertise: Biomarker discovery, target identification, multi‑omics data analysis, translational bioinformatics
